gian/cubefs
1
0
Fork 0
mirror of https://github.com/cubefs/cubefs.git synced 2026-08-02 02:00:56 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

feat(master): if crossZone true, support degrade to create partitions when only one zone available

Browse Source 
close:#22548189
Signed-off-by: true1064 <tangjingyu@oppo.com>
This commit is contained in:
true1064 2024-08-28 17:00:03 +08:00 committed by AmazingChi
parent aa716d9fc0
commit 282bb4bde2
2 changed files with 37 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
master/cluster.go
View File

@ -2020,9 +2020,10 @@ func (c *Cluster) getZoneListFromVolZoneName(vol *Vol, mediaType uint32) (zoneLi
// otherwise, return defaultReplicaNum
func (c *Cluster) decideZoneNum(vol *Vol, mediaType uint32) (zoneNum int) {
if !vol.crossZone {
log.LogInfof("[decideZoneNum] to create vol(%v), zoneName(%v) mediaType(%v), crossZone is not set so decide zoneNum: %v",
zoneNum = 1
log.LogInfof("[decideZoneNum] to create vol(%v), zoneName(%v) mediaType(%v), crossZone is not set, decide zoneNum: %v",
vol.Name, vol.zoneName, proto.MediaTypeString(mediaType), zoneNum)
return 1
return
}
specificZoneListOfMediaType := c.getZoneListFromVolZoneName(vol, mediaType)
@ -2146,6 +2147,9 @@ func (c *Cluster) getHostFromNormalZone(nodeType uint32, excludeZones []string,
excludeHosts []string, replicaNum int, zoneNumNeed int,
specifiedZoneName string, dataMediaType uint32) (hosts []string, peers []proto.Peer, err error,
) {
log.LogInfof("[getHostFromNormalZone] dataMediaType(%v) nodeType(%v) replicaNum(%v) zoneNumNeed(%v) specifiedZoneName(%v)",
proto.MediaTypeString(nodeType), nodeType, replicaNum, zoneNumNeed, specifiedZoneName)
var zonesQualified []*Zone
if replicaNum <= zoneNumNeed {
zoneNumNeed = replicaNum

43
master/topology.go
View File

@ -1361,6 +1361,10 @@ func (t *topology) getZonesOfNodeType(nodeType NodeType, dataMediaType uint32) (
return true
}
if !zone.canWriteForNode(nodeType, 1) {
return true
}
zones = append(zones, zone)
return true
})
@ -1435,22 +1439,27 @@ func (t *topology) pickUpZonesByNodeType(zones []*Zone, nodeType NodeType, dataM
// Choose the zone if it is writable and adapt to the rules for classifying zones
func (t *topology) allocZonesForNode(rsMgr *rsManager, zoneNumNeed, replicaNum int, excludeZone []string,
specialZones []*Zone, dataMediaType uint32) (zones []*Zone, err error) {
log.LogDebugf("[allocZonesForNode] NodeType(%v) zoneNumNeed(%v) replicaNum(%v) excludeZone(%v) specialZonesLen(%v) dataMediaType(%v)",
NodeTypeString(rsMgr.nodeType), zoneNumNeed, replicaNum, excludeZone, len(specialZones), proto.MediaTypeString(dataMediaType))
log.LogInfof("[allocZonesForNode] NodeType(%v) dataMediaType(%v) zoneNumNeed(%v) replicaNum(%v) excludeZone(%v) specialZonesLen(%v) ",
NodeTypeString(rsMgr.nodeType), proto.MediaTypeString(dataMediaType), zoneNumNeed, replicaNum, excludeZone, len(specialZones))
if len(t.domainExcludeZones) > 0 {
zones = t.getDomainExcludeZones()
log.LogInfof("action[allocZonesForNode] getDomainExcludeZones zones [%v]", t.domainExcludeZones)
log.LogInfof("[allocZonesForNode] getDomainExcludeZones(%v), get zoneNum: %v",
t.domainExcludeZones, len(zones))
} else if specialZones != nil && len(specialZones) > 0 {
zones = t.pickUpZonesByNodeType(specialZones, rsMgr.nodeType, dataMediaType)
zoneNumNeed = len(zones)
log.LogInfof("action[allocZonesForNode] pick up mediaType(%v) specialZones: %v",
log.LogInfof("[allocZonesForNode] pick up mediaType(%v) from specialZones, get zoneNum: %v",
proto.MediaTypeString(dataMediaType), zoneNumNeed)
} else {
// if domain enable, will not enter here
zones = t.getZonesOfNodeType(rsMgr.nodeType, dataMediaType)
log.LogInfof("[allocZonesForNode] pick up mediaType(%v) from all zone, get zoneNum: %v",
proto.MediaTypeString(dataMediaType), len(zones))
}
if len(zones) == 1 {
log.LogInfof("action[allocZonesForNode] pick up mediaType(%v) only one zone: %v",
proto.MediaTypeString(dataMediaType), zones[0].name)
return zones, nil
}
@ -1469,23 +1478,33 @@ func (t *topology) allocZonesForNode(rsMgr *rsManager, zoneNumNeed, replicaNum i
}
if zone.canWriteForNode(rsMgr.nodeType, uint8(demandWriteNodesCntPerZone)) {
log.LogDebugf("[allocZonesForNode] pick up candidate zone: %v", zone.name)
log.LogInfof("[allocZonesForNode] nodeType(%v) dataMediaType(%v), pick up candidate zone: %v",
NodeTypeString(rsMgr.nodeType), proto.MediaTypeString(dataMediaType), zone.name)
candidateZones = append(candidateZones, zone)
} else {
log.LogInfof("[allocZonesForNode] nodeType(%v) dataMediaType(%v), not enough writable node, skip zone: %v",
NodeTypeString(rsMgr.nodeType), proto.MediaTypeString(dataMediaType), zone.name)
}
}
// if across zone,candidateZones must be larger than or equal with 2,otherwise,must have a candidate zone
if (zoneNumNeed >= 2 && len(candidateZones) < 2) || len(candidateZones) < 1 {
log.LogInfof("[allocZonesForNode] nodeType(%v) dataMediaType(%v), candidate num: %v",
NodeTypeString(rsMgr.nodeType), proto.MediaTypeString(dataMediaType), len(candidateZones))
if len(candidateZones) < 1 {
if rsMgr.nodeType == DataNodeType {
err = proto.ErrNoZoneToCreateDataPartition
} else {
err = proto.ErrNoZoneToCreateMetaPartition
}
log.LogErrorf("action[allocZonesForNode],reqZoneNum[%v],candidateZones[%v],demandWriteNodes[%v],err:%v",
zoneNumNeed, len(candidateZones), demandWriteNodesCntPerZone, err.Error())
log.LogErrorf("[allocZonesForNode] nodeType(%v), dataMediaType(%v), reqZoneNum[%v], candidateZones[%v], demandWriteNodes[%v], err:%v",
NodeTypeString(rsMgr.nodeType), proto.MediaTypeString(dataMediaType), zoneNumNeed, len(candidateZones),
demandWriteNodesCntPerZone, err.Error())
return nil, err
}
if zoneNumNeed >= 2 && len(candidateZones) == 1 {
log.LogWarnf("[allocZonesForNode] nodeType(%v), dataMediaType(%v), demandWriteNodes[%v], reqZoneNum is [%v] but only one candidateZone",
NodeTypeString(rsMgr.nodeType), proto.MediaTypeString(dataMediaType), demandWriteNodesCntPerZone, zoneNumNeed)
}
zones = candidateZones
err = nil
return
@ -1801,7 +1820,7 @@ func (zone *Zone) allocNodeSetForMetaNode(excludeNodeSets []uint64, replicaNum u
return ns, nil
}
func (zone *Zone) canWriteForNode(nodeType NodeType, replicaNum uint8) (can bool) {
func (zone *Zone) canWriteForNode(nodeType NodeType, demandWriteNodesCntPerZone uint8) (can bool) {
zone.RLock()
defer zone.RUnlock()
@ -1823,7 +1842,7 @@ func (zone *Zone) canWriteForNode(nodeType NodeType, replicaNum uint8) (can bool
if node.IsActiveNode() && node.IsWriteAble() {
leastAlive++
}
if leastAlive >= replicaNum {
if leastAlive >= demandWriteNodesCntPerZone {
log.LogDebugf("[canWriteForNode] canWrite: nodeId(%v) addr(%v) zone(%v) nodeType(%v)",
node.GetID(), node.GetAddr(), zone.name, NodeTypeString(nodeType))
can = true